Originally Posted by Kris Deering
Dolby Vision would not only know the peek white level of the display for proper tone mapping, but also the measured black level as well. It also has all the data per frame so would be ideal for a dynamic contrast system (laser or iris) for adjusting on a frame by frame basis the APL.
So I get where static metadata is useful (in theory), you can know, for example, that you don't have to try to retain all 10,000 nits in a calibration, when no content in a given movie will go over 1100. Lumagen is doing this with their Intensity Mapping function, automatically adjusting the bounds, but this causes problems when the metadata isn't correct. However I don't see how frame-by-frame helps, the min and max luminance can be determined easily on a per frame basis in real time, without metadata (though that makes it easier). Average level is relatively easy to figure out too. What else is included in the dynamic metadata? Mastering monitor min/max level should be static for an entire film right? And the 2084 EOTF is fixed, not relative, so there shouldn't be anything like gamma adjustments. One thing I could see being in there is specifying what the "nominal" white point in the frame is, which could be useful in knowing what not to clip or roll off, but as we've seen with the static metadata, it seems like that just adds another order of magnitude potential for errors.
It seems to me that per-frame metadata may make things a bit more efficient, save some processing on the display side, but it also opens up a path for a whole nother level of FUBAR-ness. For example if you tried to adjust the mapping function on a per-frame level, that would be likely to have terrible results, like the dynamic gamma feature on Lumagen/Gennum processors, you could end up with pumping in the highlights or dark level whenever the peak white or black in a scene changes.
I realize there are ways to tell something these levels, but you have to remember mass implementation can't rely on some calibrator telling it something at some point. Dolby would want that mode to kick on from minute 1 of use for anything input that is encoded as Dolby Vision, and since it doesn't know those variables (and the variation from unit to unit could be MASSIVE here) it puts them in a tough spot. Most Dolby Vision displays allow ZERO adjustment once they go in that mode, and I could see them wanting to do the same thing here. Again, it wouldn't surprise me in the slightest if we never see DV support in projectors at all.
Yup, like I said, I understand why Dolby wants to keep control, from a political standpoint. I'm really more curious how dynamic metadata (one of the key features of DV) is really supposed to help things.